Twelve trials (n=5,060 participants) from nine countries met the inclusion criteria and they were all published in English. Eight trials compared amiodarone with a rate-control drug and four trials compared it with a placebo. Concealment of allocation was adequate in eight trials, inadequate in one trial, and unclear in two trials. Six trials were double blind, two were single blind, and in four trials the outcome assessment was not blind. All the trials reported an intention-to-treat analysis.
Where reported, the duration of follow-up ranged from six to 60 months. There was statistically significant heterogeneity between trials for all secondary outcomes and the funnel plot showed the possible presence of publication bias.
There was no statistically significant difference between amiodarone and placebo or a rate control drug, in long-term mortality: 4.7 versus 3.9 per 100 patient-years (12 trials, n=5,060); one trial received a weight of 71.46%. Amiodarone was more effective compared with placebo or a rate control drug in achieving sinus rhythm: 21.3 versus 9.2 per 100 patient-years (RR 3.2, 95% CI 1.9 to 5.5; 11 trials, n=5,044). The incidence of hospitalisation did not increase with amiodarone treatment compared with placebo or a rate control drug (five trials, n=2,932).
The risk of withdrawal from the trial was significantly greater in the amiodarone group compared with placebo or a rate control drug: 10.7 versus 1.9 per 100 patient-years (RR 3.0, 95% CI 1.4 to 6.2; nine trials, n=3,176). These findings did not change in any of the sensitivity analyses.